This thesis addresses the problem of generating a range of natural sounding pitchudcontours for speech synthesis to convey the specific meanings of different intonationudpatterns.udWhere other models can synthesise intonation adequately for short sentences,udlonger sentences often sound unnatural as phrasing is only really considered atudthe sentence level. We build models within a framework of prosodic structureudderived from the linguistic analysis of a corpus of speech. We show that the useudof appropriate prosodic structure allows us to produce better contours for longerudsentences and allows us to capture the original style of the corpus. The resultingudmodel is also sufficiently flexible to be adapted to suitable styles for use in otheruddomains.udTo convey specific meanings we need to be able to generate different accentudtypes. We find that the infrequency of some accent and boundary types makesudthem hard to model from the corpus alone. We address this issue by developinguda model which allows us to isolate the parameters which control specific accentudtype shapes, so that we can reestimate these parameters based on other data.
展开▼
